2 Things Marketers Must Know about Data

2 Things Marketers Must Know about Data:

Most of us are perfectly content not knowing how everything works, we push a button, turn a handle, move a mouse or scan a card and voila! The things we want done are done. But there are times that knowing a thing or two about the mechanics saves you great trouble in the future in the forms of mistakes avoided and opportunities salvaged.

In our most recent survey on the future of marketing, 80% of B2B and 94% of B2C marketers tell us they expect analytics to inform their day-to-day business decision making in the near future (more on this in later posts). While you don’t need to become a quant person overnight, it’s helpful to understand a few key statistical concepts. Here are two things that every marketer should know about data:

Sample size. The sample size gives you a sense whether or not you have that “critical mass” of people (customers, for most purposes) on which to base your conclusions. When dealing with sample sizes, watch out for a few things:

• Too small. A small sample size will result in “really?!” glances from your audience unless you have a very good reason (i.e. there are 10 pandas in the world and 9 are in my dataset).
• Too big. If you already have 5000 people, you don’t need another 2000. It’s the Law of Large Numbers – i.e. if you’re large, being larger doesn’t make you much different.
• Not representative. Marketers prefer larger samples if only to feel buttressed against hostile rebuttals but statisticians would direct you to check for “representativeness” — how well the composition of your sample fits the needs of the project at hand. If your online customers have different buying preferences compared to your brick-and-motar customers, I’d make sure to get the right names out of the CRM system when testing the efficacy of an in-store campaign before worrying about response rates. The same goes with any substantial differentiating factors like race, gender and age.

Significance level. Significance is an econometrics term; basically, it’s a measure of to what degree a particular variable is responsible for reaching an outcome. For instance, price might be a more significant variable when it comes to market share than, say, store placement. When dealing with significance, here’s what to watch for:

• Not really causal. When a variable is suspected of causing an outcome we don’t really know whether it’s guilty as charged, acting as cover for another factor, or whether the causation runs the opposite way. Corporate culture, for example, could show up as a “driver” of business outcomes when all that’s going on is people attributing openness and innovation to their company when the business is doing well. Phil Rosenzweig covered this aptly in “The Halo Effect”.
• Insignificance might matter. If significance is statistically prominent, then the insignificance of something you thought would matter should capture your attention like an awkward silence. This is your chance to debunk some conventional wisdom and find potential areas to innovate/change. Surprised that simple cheerleading on Facebook didn’t drive loyalty? Time to ask fans what would worth their while.

Keeping these two concepts in mind will help a lot of marketers – particularly folks at the mid and junior levels – conversant in data, and will help them apply statistical concepts to their work. MLC members, we’re curious – how are you training non-quant staff in statistical concepts? Let us know in comments.